Cow's milk protein allergy masked by suspected necrotizing enterocolitis: diagnostic delay and serial eosinophil...

This study aimed to investigate the diagnostic utility of serial peripheral blood eosinophil percentage (PBEP) monitoring in differentiating cow's milk protein allergy (CMPA) from necrotizing enterocolitis (NEC) in neonates, and to quantify the diagnostic delay associated with a preceding NEC diagnosis in neonates subsequently diagnosed with CMPA.

This retrospective cohort study was conducted at a tertiary neonatal intensive care unit (NICU) between 2015 and 2025. A total of 112 neonates were enrolled and divided into three groups: Group 1 (NEC + CMPA, n = 39), Group 2 (CMPA only, n = 33), and Group 3 (NEC only, n = 40). PBEP was assessed at two time points, initial admission and at the time of confirmed diagnosis, for all groups. The magnitude of eosinophil change (ΔPBEP) was calculated for each patient. Receiver operating characteristic (ROC) curve analysis with Youden Index-based cutoff determination was performed to evaluate the discriminative performance of PBEP. Multinomial logistic regression analysis was applied to identify independent determinants of group membership.

Initial PBEP was comparable across all three groups (p = 0.685). At follow-up, PBEP increased markedly in Group 1 (15.57 ± 6.84%) and Group 2 (12.43 ± 6.93%), while remaining low in Group 3 (2.98 ± 1.62%) (p < 0.001). ΔPBEP was significantly greater in Group 1 and Group 2 compared with Group 3 (p < 0.001). ROC analysis demonstrated poor discriminative performance for initial PBEP (AUC = 0.463), whereas follow-up PBEP showed excellent diagnostic accuracy (AUC = 0.986; optimal cutoff: ≥5.20%; sensitivity: 97.2%; specificity: 92.5%; PPV: 95.9%; NPV: 94.9%). Diagnostic delay was significantly longer in Group 1 (15.2 ± 13.6 days) compared with Group 2 (7.0 ± 9.9 days) (p < 0.001). Multinomial logistic regression identified the absence of sepsis as the only independent determinant of membership in Group 2 (OR = 4.319; 95% CI: 1.217-15.323; p = 0.024), and ΔPBEP as the only independent determinant distinguishing Group 3 from Group 1 (OR = 0.047; 95% CI: 0.002-0.945; p = 0.046).

Serial PBEP monitoring may serve as a practical, noninvasive, and readily available biomarker for differentiating CMPA from NEC in the NICU. While initial PBEP lacks discriminative value, follow-up PBEP demonstrates excellent diagnostic accuracy. A preceding NEC diagnosis may be associated with a substantial diagnostic delay, highlighting the importance of considering CMPA in neonates with recurrent suspected NEC episodes, particularly when gastrointestinal symptoms persist or recur despite standard NEC treatment.